Reimagining the Software Crisis:
Myhrvold began his speech with the intent to discuss the future of software over the next half-century, but reframed this talk under the title “Software, the crisis continues,” implying the ongoing challenges in the software industry.

His perspectives are primarily shaped by his experiences as an unrepentant programmer and a theoretical physicist.

Creation of the Universe: A Cosmologist’s Perspective:
He humorously attempted a brief recap of the universe’s creation, from its start as a quantum gravitational fluctuation to the formation of the Santa Z convention center.

He acknowledged the oversimplification of this process, which was intended to establish his bold approach to discussing complex topics.

The Evolution of Information:
The history of information is a key subject for any discussion on software, according to Myhrvold.

He stated that before the invention of writing, information as we understand it didn’t exist. The best ideas couldn’t be effectively passed on and were lost.

The invention of writing was a significant revolution in information. It started to make a significant impact when the printing press was invented by Johann Gutenberg. This began an extensive information revolution, paving the way for many subsequent improvements.

Computers and Information Revolution:
Myhrvold highlighted that the invention of the computer fundamentally changed how humans store, manage, and manipulate information. However, initial access was restricted to few, similar to the beginnings of writing.

The advent of the microprocessor and networking democratized computing, and the real essence of the internet lies in connecting people, not just computers.

He predicted that the ongoing information revolution, spurred by cheap, ubiquitous information distributed through these mediums, will impact society as profoundly as the industrial revolution.

Moore’s Law as a Driver:
The technical driver of this revolution, according to Myhrvold, is Moore’s Law, which predicts a doubling of computer processing power every two years.

Over the past 20-25 years, we have seen a factor of a million improvement in computing power, and Myhrvold believes this trend is likely to continue for another 20-40 years.

He expressed confidence in this projection due to his understanding of the underlying physics, the fact that this is primarily a hardware problem, and the enormous scale of potential improvements.

The Impact of Technological Progress:
Myhrvold illustrated the potential impact of such progress by positing that in 20 years, a task that takes a contemporary computer a year to complete could be done in 30 seconds, while in 40 years, the same task could be done in 30 seconds by a machine that would take a contemporary computer a million years.

He highlighted the crucial role of software in harnessing this power to provide useful services and functions for people.

Moore’s Law and Software Economics:
Myhrvold presented a chart illustrating the exponential growth of the number of bytes you get for a dollar, which aligns with Moore’s Law.

He also shared a graph showing the stock price of a certain software company growing at a similar exponential rate.

Upon comparison, he found the growth rate of the software company’s stock price almost exactly matches the delivered performance rate predicted by Moore’s Law.

Based on these correlations, Myhrvold inferred that Moore’s Law has direct implications for the software industry’s economic growth.

Nathan’s Laws of Software:
Myhrvold introduces his “laws” of software, which offer insights into the nature and evolution of software development. These laws reflect his observations and experience in the field of software.

Software is a Gas:
Myhrvold’s first law posits that software, like gas, expands to fit its container. This is evidenced by the exponential growth in the size of software like Windows NT and Microsoft Word. The lines of code used in these software have increased significantly over time. This expansion is driven by the continuous addition of features and capabilities to meet user expectations.

Software Expansion and Moore’s Law:
The second law suggests that software grows until it is limited by Moore’s Law, which states that the number of transistors on a chip doubles approximately every two years. The expansion of software eventually hits the limitations of the available hardware. However, software has been observed to grow even faster than Moore’s law, frequently maxing out the potential of current hardware just before the next generation becomes available.

The Interdependency of Software and Hardware:
Myhrvold’s third law suggests that software growth makes Moore’s Law possible. People buy faster and more powerful machines because the software has outgrown the capabilities of their old hardware. Thus, the continuous demand for advanced hardware is driven by the relentless growth of software.

Human Ambition and Expectation Drive Software Growth:
The fourth law states that software is limited only by human ambition and expectation, implying almost no limitation. The pursuit of perfection, development of new algorithms, discovery of new applications, and changing definitions of what’s considered “cool” all contribute to continuous software development.

Impact of Gaming on Software and Hardware:
Myhrvold observes that games push the boundaries of hardware and software, often requiring the latest, most powerful machines to function properly. These games provide users with cutting-edge experiences, which drives further demand for advanced hardware and software.

Software Crisis in History:
Nathan Myhrvold, a board member of the Institute for Advanced Study, shares anecdotes from his dinner with John Von Neumann’s assistant, where she recounted Neumann’s frustration with the software aspect of his project, despite his monumental contributions to computing theory. This serves as an early example of the so-called ‘software crisis’.

The term ‘software crisis’ has been recurrent in discourse since the 1950s, characterizing a constant state of difficulty and challenge in software development.

Understanding the ‘Software Crisis’:
This ‘crisis’ is tied to the ever-increasing complexity in software engineering, fueled by ambition, expectations, and creativity. As soon as solutions to current issues are developed, they are absorbed and new, more complex problems arise, leading to a perception of a perpetual crisis.

The issue is not with the development or maturity of software engineering as a discipline, but with the nature of software itself. Unlike other engineering fields, software engineering is essentially a continuous process of problem-solving, making it fundamentally different and perhaps more challenging.

Software Progress and Productivity:
Myhrvold observes that every new approach or technology – high-level languages, object-oriented programming, software components – proposed as a solution to the ‘software crisis’ gets rapidly absorbed and used to push boundaries further.

He notes a significant increase in software productivity over time. For instance, at Microsoft, productivity has increased by a factor of 10 over the past decade, largely due to the use of these ‘solutions’.

The Dynamics of Software Requirements:
Myhrvold argues that software will never be ‘easy’ unless we reduce our expectations and requirements. There is always a push to increase complexity and functionality in software development.

He presents the idea that while stability is desirable for some software (e.g., ABS in cars), the drive for innovation and ‘cool’ new features in other areas (e.g., gaming) leads to a constant push of boundaries, contributing to the perpetual ‘software crisis’.

The Future of Software: Opportunities and Challenges:
Myhrvold posits that the future of software will be marked by great strides in creative and practical applications, like creating synthetic actors for movies. However, these advancements will also become more widely accessible, presenting both opportunities and challenges.

The rapidly growing computing power and its implications necessitate forward-thinking and an effort to stay ahead of this exponential wave.

Emergence of Genetic Programming:
Nathan Myhrvold discusses the advent of genetic programming as an innovative approach in software development. This technique involves evolving programs through simulations of evolution, a concept he likens to “software husbandry.”

He shares his experience in developing a small-scale genetic program in 1987, which has been used in almost all Microsoft products since Windows 2.0.

Comparing Software and Biological Complexity:
Myhrvold introduces an interesting perspective on the comparative complexity of software and biological information. He uses an illustrative example comparing the data size of Madonna’s movie ‘Evita’ and her DNA, revealing that the movie contains more data.

He further explains that only about a quarter percent of the DNA of two individuals differ, meaning a significant part of the genome could be compressed.

A surprising outcome of his analysis is that all the genetic information that makes an individual unique could potentially fit on a floppy disk.

Genetic Diversity vs. Artificial Data:
In terms of genetic diversity, Myhrvold argues that the genomes of all humans on Earth could be stored on a server with about 3.7 terabytes of storage, pre-compression. Post-compression, this could be reduced to less than a terabyte, or the size of a large website.

Extending this analogy, he suggests that the genomes of all animals on Earth might be roughly equivalent to the size of the web.

He clarifies that this comparison does not diminish the value of biology, but it is surprising that the information produced by 3 billion years of evolution could be comparable in size to the information humans have generated on the web in a few years.

The Computational Simplicity of Biological Information:
Myhrvold argues that, despite the awe-inspiring capabilities of molecular biology and genetic technology, computationally they’re not that complex.

He suggests that, given the development of technology, we’ll soon be able to compute all possible combinations of DNA sequences. This is because the capability of our computational systems to handle such information continues to grow.

Human Brain as Ultimate Computer: Nathan Myhrvold begins his talk by framing the human brain as what we currently consider the ultimate computer. He acknowledges the 1960s and 1970s optimism around artificial intelligence, suggesting that people believed AI could potentially equal or surpass the human brain without necessarily understanding how it functions.

Forecast of Artificial Intelligence: Despite prior disappointments, Myhrvold remains optimistic about the future of AI. He anticipates artificial intelligence surpassing the intellectual capabilities of a dog within 15 years and eventually reaching the human intelligence level, potentially within a 20 to 30 year period.

Limitations of Current AI: While he envisions a future where AI can match human intelligence, Myhrvold is clear that current AI capabilities are limited. They do not match even those of his dog in terms of speech recognition, simple tasks, reasoning, and vision.

The “Boot Time” of Humans and AI: Myhrvold uses an analogy between the process of “booting up” a computer and the human learning period, which he jokingly states takes approximately 20 years. This humorously illustrates the complexity of programming AI to mimic human intelligence and the potential challenges of “teaching” computers as we do humans.

Neuroscientific Approach to AI: He explores the potential of a neuroscientific approach to AI, proposing the concept of emulating human brain activity through advanced hardware and software. This could involve methods such as scanning a human brain to capture its structure and function, though he acknowledges this concept is still speculative.

Future Operating Systems and Software: Myhrvold further speculates on what operating systems and software might look like in a future where human cognition can be uploaded to a computer system. He predicts the need for an operating system that can house multiple “uploaded” human personalities, allow for multitasking and provide virtual memory. He humorously compares the features of Windows 2.0 to those of a hypothetical Windows 2047, which he imagines will have to deal with backward compatibility issues with “meat-based” humans.

Transcending Physical Limitations: Ultimately, Myhrvold’s talk focuses on a future where artificial intelligence and human consciousness intertwine. He envisions a reality in which humans might transcend their physical limitations to exist as software, further blurring the lines between human intelligence and artificial intelligence.